Needing more than a spark test?

Getting a copper entirely around the inductor, but in a way that avoids a shorted turn, will stop any antenna-type E-Field. I have done this using a ground plane under all of a small PSU circuit, and bringing the metal shield sides up from it, but when completing the "box" over the top, I used two "lids", one folded over the top of the other, but with a thin sticky tape between them. The remaining two ends got the same treatment. In appearance, it was a untidy homemade bodge-up, but it made a Faraday "cage". The overlapping sufaces with sticky tape between them completed the Faraday cage by the capacitance acting as an AC short-circuit. It's very difficult for the field trapped in the thin dielectric between two coppers that are DC shorted together, to make any external electric near field to affect anything, much less to start any real farfield EM propagation.

Alternating magnetic near-fields cannot escape a truly shorted Faraday cage (shorted turn effect), but they can escape far and wide if the cage is not truly shorted, like the inductor shield I describe. The magnetic near field can inductively couple to traces, and add in their noise. Fascinating (to me) is how they can still get a true RF propagation going, including generating a new electric far-field. The electric near-field component may be trapped in the shield overlap, but the magnetic component can still get out and make a true RF radiation happen. That is how direction-finding shielded Goniometer's, and communication magloops work as antennas.

At the dimensions of the PCB, it's all so close in terms of wavelengths that only near-fields can be coupling this noise into the traces. Moving the entire switcher away, and having a a piece of mu-metal, and maybe some ferrite in the design of the sense head assembly, should approach having the noise reduce to only that of thermal, diode, and amplifier front-end that we already know about.

I have to agree. It is simply obvious that the solution to switcher noise is to move it away, and filter the noise. For the purposes of your experiment, it may be faster and easier just to use a battery or two hooked up. You know the final circuit will have the PSU live elsewhere (and be a clean one) anyway!

Thanks for the seasonal wishes, and likewise for yourself.
 
Some good comments there! Over the years I've had to deal with unwanted contributions from my measurement system (most troublesome, vibration-prone coax). This is just more of the same and yet another reason to start with a testbed kind of setup. It's just as valuable to determine what doesn't work as what does......
 
I get that :) I just want a better hit rate than this guy..

“I have not failed. I've just found 10,000 ways that won't work.”
― Thomas A. Edison
 
Would your external power supply be linear instead of switching?
 
rwm said:
Would your external power supply be linear instead of switching?
Click to expand...
Yes indeed! I'm thinking about a separate power supply board to produce all the required voltages. Linear regulator IC's in the 78xx/79xx family are cheap and easy to wire up, even with my old eyes. The main thing is to remember that Vin and Gnd are swapped when you go from a positive regulator to a negative regulator :) The circuitry draws so little current I don't think I will even need heat sinks on them.
 
rwm said:
Would your external power supply be linear instead of switching?
Click to expand...
Good question!
In theory, it could be either. In practice, it might well be switcher-derived, then followed by linear noise clean-up.

Part of "clean-up", in addition to the usual noise filtering, can include a final linear low drop-out regulator stage designed to have the output only a little below the switcher-type input. This final regulator is itself low noise, and has very high bandwidth, fast enough to "follow" all the noise, and regulate it away. Very important is how clean the reference is, and how clean is the bottom end common mode return route. If the 0V for the reference just casually connects to the switcher 0V, it can simply import a mad racket onto the thing you are trying to keep clean.

Paying attention to the return current paths back to the common star point, and not allowing the linear regulator reference to be bounced around by switcher common mode noise, gives a clean instrumentation grade supply. The same attention has to apply to A/D converters, and their references. The bad scene there is the low-order bits being jangled around, sampling random noise on top of the signal you want.

The tricks need not be expensive, nor too elaborate. If the amplifier does not need big currents, then something as simple as putting two resistors into it's energy route, and shunting them with a capacitor for energy storage, in parallel with a smaller value high frequency capacitor to deliver transients, can wipe out quite severe noise. I prefer to have the noise as low as possible at the start, and then include the clean-up anyway.
 
homebrewed said:
Yes indeed! I'm thinking about a separate power supply board to produce all the required voltages. Linear regulator IC's in the 78xx/79xx family are cheap and easy to wire up, even with my old eyes. The main thing is to remember that Vin and Gnd are swapped when you go from a positive regulator to a negative regulator :) The circuitry draws so little current I don't think I will even need heat sinks on them.
Click to expand...
78xx and 79xx certainly are cheap and plentiful.
Cautions. Any that get to their voltage by using a diode in their common return, may have noise issues. Linear they are, so they at least do not have switcher flyback noise. I know the 78xx series has been going since 1972, and is still the hobbyist's favourite. There have since been some sporting the same number in their code, that have had some improvements. There are now low cost low noise regulators that may be a better choice.

Mark is right about the low current and not needing heatsinks. Earlier on in this thread, one of our simulation circuits had a low noise op-amp provide the 0V regulated return to be forced exactly halfway between the offered supply voltage, so generating a dual-rail power supply to allow inverted waveform gain stages.
 
Like numerous amplifiers claimed to be "low noise" but aren't, I've found quite a few "low noise" regulators that actually don't look particularly good, with noise voltages in the hundreds of micro-volts (according to their data sheets). Thank goodness decent amplifiers have good power supply rejection! That said, poor design around good amplifiers equals a waste of money.
 
Wrapping the PocketGeiger with a grounded copper-foil shield didn't do much as far as getting rid of the switcher noise. I suspected it wouldn't, but it was worth trying.

The noise level from that source is high enough to boot the thing out as a viable XRF detector so there's no question it has to be addressed. The DIN connector I'm using has one spare pin so that's where the diode bias voltage will come in.

There was one interesting positive result from this experiment. The design of my enclosure allows me to remove the Am241 sources without messing up the electrical shielding, so I did that just to avoid exposure to the (admittedly low) x-ray dose. As a result, the background pulse count went down to zero. Apparently there are enough x-rays making it through the .25" thick aluminum for the detector to pick them up. So before I re-install my X-ray source "plate", I will back it with a layer of lead shielding.

So the executive summary is: No switchers in the box!!! And more shielding between the Am241 sources and detector is needed. Not all negative results, per Graham's Edison quote.
 
So there you have it.
X-rays can just waltz through a quarter inch of aluminum! That does surprise me.
It well explains why those photomultiplier tubes can have an aluminum wall construction by the photocathode.

Re: Only one pin spare. OK - try it. It may be all we need.
Here my thoughts were that the pin's purpose is the diode bias voltage. The current is tiny, and competes with the currents caused by the X-ray photons. If that current return path has to share with larger currents from other circuitry on the PCB, then the clean bias supply carries noise voltages added in from what it collected in by the resistances, and inductances in that path, and they then get amplified.

This is a purist design point. At present, II also have only one pin. In my scheme, the ADC is not on that board. Instead, it gets high level amplified low impedance analogs sent to it down via the cable. The amplifiers and stuff on the board are low current things, unlikely to build much unwanteds, except across lead inductance. These are pulses. They make transient currents. That is where I reasoned the bias supply should be clean, noise free, isolated, and absolute. It can bias the diode, and that is all. This is why I considered using a small battery.

Forgive that my thoughts are a bit like kibbutz from the sidelines for now. I am extremely curious as to the state of these real signals from real X-Rays. For average HM users, this must seem like tortuous obsession nerd stuff, but I consider where you are at is real progress! :)

Happy Christmas all!
 
You must log in or register to reply here.
Back
Top